diff --git a/codes/models/steps/losses.py b/codes/models/steps/losses.py
index d367d318..69735a89 100644
--- a/codes/models/steps/losses.py
+++ b/codes/models/steps/losses.py
@@ -51,6 +51,8 @@ def get_basic_criterion_for_name(name, device):
         return nn.L1Loss().to(device)
     elif name == 'l2':
         return nn.MSELoss().to(device)
+    elif name == 'cosine':
+        return nn.CosineEmbeddingLoss().to(device)
     else:
         raise NotImplementedError
 
@@ -79,7 +81,10 @@ class FeatureLoss(ConfigurableLoss):
         with torch.no_grad():
             logits_real = self.netF(state[self.opt['real']])
         logits_fake = self.netF(state[self.opt['fake']])
-        return self.criterion(logits_fake, logits_real)
+        if self.opt['criterion'] == 'cosine':
+            return self.criterion(logits_fake, logits_real, 1)
+        else:
+            return self.criterion(logits_fake, logits_real)
 
 
 # Special form of feature loss which first computes the feature embedding for the truth space, then uses a second